U.S. patent application number 13/137119 was filed with the patent office on 2012-03-01 for movable robot.
This patent application is currently assigned to TOYOTA JIDOSHA KABUSHIKI KAISHA. Invention is credited to Haeyeon Lee, Yuichiro Nakajima, Hideki Nomura.
Application Number | 20120053727 13/137119 |
Document ID | / |
Family ID | 38371597 |
Filed Date | 2012-03-01 |
United States Patent
Application |
20120053727 |
Kind Code |
A1 |
Nakajima; Yuichiro ; et
al. |
March 1, 2012 |
Movable robot
Abstract
A technique to wholly recognize the surrounding environment may
be provided by excluding unknown environment which arises due to
parts of a body of a robot hindering the sight of the robot during
operations. The robot of the present invention is provided with a
body trunk including head and torso, at least one connected member
that is connected to the body trunk by a joint in which a driving
mechanism is provided, a body trunk side camera that is arranged on
the body trunk, and a connected member side camera that is arranged
on the connected member. Further, the robot is provided with a
composite image creation unit that creates composite image of a
body trunk side image taken by the body trunk side camera and a
connected member side image taken by the connected member side
camera, such that a part of the body trunk side image is replaced
with a part of the connected member side image so as to exclude the
connected member from the body trunk side image.
Inventors: |
Nakajima; Yuichiro;
(Toyota-shi, JP) ; Lee; Haeyeon; (Toyota-shi,
JP) ; Nomura; Hideki; (Nagoya-shi, JP) |
Assignee: |
TOYOTA JIDOSHA KABUSHIKI
KAISHA
|
Family ID: |
38371597 |
Appl. No.: |
13/137119 |
Filed: |
July 21, 2011 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
12279561 |
Aug 15, 2008 |
8010232 |
|
|
PCT/JP2007/052753 |
Feb 15, 2007 |
|
|
|
13137119 |
|
|
|
|
Current U.S.
Class: |
700/259 ; 901/1;
901/47; 901/9 |
Current CPC
Class: |
G06T 2200/32 20130101;
G01B 11/03 20130101; B25J 5/007 20130101; G05D 1/0251 20130101;
G05D 1/0274 20130101 |
Class at
Publication: |
700/259 ; 901/1;
901/9; 901/47 |
International
Class: |
B25J 13/08 20060101
B25J013/08; B25J 19/04 20060101 B25J019/04 |
Foreign Application Data
Date |
Code |
Application Number |
Feb 17, 2006 |
JP |
2006-041323 |
Claims
1-5. (canceled)
6. A movable robot, comprising: a body trunk; a connected member
that is connected to the body trunk by a joint in which a driving
mechanism is provided; a plurality of body trunk side cameras that
is arranged on the body trunk; a connected member side camera that
is arranged on the connected member; a first calculating unit that
calculates first relative positional relationship, with respect to
the movable robot, of one or more objects existing within a
photographed scope based on at least two body trunk side images
taken by the plurality of body trunk side cameras; and a second
calculating unit that calculates second relative positional
relationship, with respect to the movable robot, of one or more
objects existing within a photographed scope based on a body trunk
side image taken by at least one of the plurality of body trunk
side cameras and a connected member side image taken by the
connected member side camera; wherein the second calculating unit
selects a body trunk side image in which portion of the connected
member being photographed is the smallest to be used in the
calculation of the second relative positional relationship.
7-8. (canceled)
Description
CROSS REFERENCE TO RELATED APPLICATION
[0001] This application claims priority to Japanese Patent
Application No. 2006-041323, filed on Feb. 17, 2006, the contents
of which are hereby incorporated into this specification by
reference.
TECHNICAL FIELD
[0002] The present invention relates to a robot that moves with
self-driven power. Particularly, the present invention relates to a
robot that moves by recognizing the surrounding environment using
cameras.
BACKGROUND ART
[0003] Robots having a connected member that is connected to a body
trunk via a joint that includes a driving mechanism are being
developed. Among the robots of this type, there is a type of robot
that creates an environmental map from environmental images that
are taken by a body trunk-side camera arranged on the body trunk,
and, based on the acquired environmental map, draws up a plan of
its moving course to the aimed position. In the technique disclosed
in the Japanese Patent Application Publication No. 2005-92820,
images are taken by a camera that is arranged on a head of a robot,
a disparity image or a distance image is calculated from the
acquired image, planar parameters are computed from the acquired
disparity image or distance image, and then creating an
environmental map by abstracting a plurality of planes including
the floorage space based on the acquired planar parameters. The
robot recognizes the areas within which the robot is able to move
from the environmental map, and plans its moving course based on
the recognized areas.
DISCLOSURE OF INVENTION
[0004] With a robot having a connected member that is connected to
a body trunk via a joint that includes a driving mechanism, there
may be a case where part of the surrounding objects cannot be
photographed by the robot's camera depending on the position of
connected member such as an arm, leg or the like, because a part of
the range of camera vision is interrupted by such connected member.
Thus, there may be a case where an unrecognizable environment
exists behind the connected member.
[0005] If such unrecognizable environment exists behind the
connected member, the moving course of the body trunk may become
difficult to plan. Moreover, the moving course of the connected
member such as the arm or leg cannot be planned. Especially in
planning a path to move the connected member to a target position,
there may often be a case where an undesirable situation that the
target position cannot be recognized due to the fact that the
target position is located within the unrecognizable
environment.
[0006] With the employment of the conventional technique, every
time an unrecognizable environment arises, extra processes of
moving the body trunk or the connected member to another position,
then photograph the unrecognized environment, and re-determine the
moving path after the unrecognized environment becomes "recognized"
environment. The conventional technique, in addition to requiring
aforesaid superfluous operations, cannot be effectively employed to
robots which move their body trunks in a continuous motion.
Furthermore, the conventional technique cannot be effectively
employed by robots in an environment in which the surrounding
objects are moving.
[0007] The present invention aims to realize a movable robot that
may prevent the occurrence of unrecognizable environment.
[0008] The present invention provides camera on the connected
member so that the area that are interrupted by the connected
member from the view of the camera on the body trunk may be thereby
photographed. Undesirable condition in which some parts of the
environment cannot be photographed due to the existence of the
connected member can thus be prevented.
[0009] The robot created by the present invention comprises a body
trunk, a connected member that is connected to the body trunk by a
joint in which a driving mechanism is provided, a body trunk side
camera arranged on the body trunk, a connected member side camera
arranged on the connected member and a composite image creation
unit. The composite image creation unit creates a composite image
of a body trunk side image taken by the body trunk side camera and
a connected member side image taken by the connected member side
camera, such that a part of the body trunk side image is replaced
with a part of the connected member side image so as to exclude the
connected member from the body trunk side image.
[0010] The aforesaid robot is able to achieve an environmental
image with a wide range of view by utilizing the body trunk side
image taken by the body trunk side camera. However, there may be a
case where the connected member is included within the body trunk
side image. In such a case, unrecognizable environment whose
environment image cannot be achieved for being hidden by the
connected member from the body trunk side camera may reside within
the said body trunk side image.
[0011] Even in such a case, the aforesaid robot comprises the
composite image creation unit that creates a composite image by
composing the connected member side image into the body trunk side
image. Hence, the area within the body trunk side image in which
the connected member is photographed and remaining as an
unrecognizable environment because of the connected member can be
replaced with a corresponding image that is taken from the
connected member side image. This results in no unrecognizable
environment remaining within the created composite image.
[0012] The aforesaid robot may compute the moving course of its
body trunk or connected member by utilizing the composite image
having no unrecognizable environment. Superfluous operations which
were conventionally required in order to get rid of the
unrecognizable environment can thus be omitted. Furthermore, even
in an environment in which the surrounding objects moves,
appropriate moving path of the robot may be timely computed.
[0013] In another alternative embodiment of the robot of the
present invention, the composite image creation unit may further
comprise a first specifying unit that specifies the area in which
the connected member is photographed within the body trunk side
image, a second specifying unit that specifies corresponding area
within the connected member side image that corresponds to the area
of the body trunk side image specified by the first specifying
unit, and a replacing unit that replaces the area of the body trunk
side image specified by the first specifying unit with the
corresponding area of the connected member side image.
[0014] As for the configuration of a member that specifies the area
in which the connected member is photographed within the body trunk
side image, the types of configuration may roughly be categorized
into two types. One type of configuration may employ the positional
relationship of the connected member with respect to the body trunk
side camera (whose relationship is known to the robot) in the
calculation of an image area in which the connected member is
assumed to be photographed. Another type of configuration may
extract a part of the body trunk side image that corresponds to the
shape or contour of the connected portion (whose shape is known to
the robot) by executing image processing on the body trunk side
image, and then specify the area in which the connected member is
photographed. Either configuration may be employed.
[0015] In the connected member side image, an area which would have
been photographed by the body trunk side camera if the connected
member had not been within the sight range of the body trunk side
camera is photographed. The configuration that specifies such an
area may also be categorized roughly into two types.
[0016] One type of configuration may employ the positional
relationship of the connected member with respect to the body trunk
side camera (whose relationship is known to the robot) and the
positional relationship of the connected member side camera with
respect to the connected member in the calculation of an image area
which would have been photographed by the body trunk side camera
had the connected member not been hindering the sight. Such
configuration may be effective especially in the case where
information of lateral distance in the depthwise direction is known
to the robot. For example, in a case where a self-driven movable
robot operates switches arranged on a wall, if the position of the
robot with respect to the wall is known to the robot, then the
scope of which to photograph the wall by the body trunk side camera
may be calculated. Furthermore, if the position of the connected
member with respect to the body trunk is known to the robot, then
the scope of the wall of which cannot be photographed due to the
existence of the connected member may be calculated. In a case
where the position of the connected member with respect to the body
trunk and the position of the connected member side camera with
respect to the connected member are given, an area within the
connected member side image that corresponds to the scope of the
wall that cannot be photographed from the body trunk side camera
for being located behind the connected member may be calculated.
Furthermore, if the area in which the connected member is
photographed within the body trunk side camera can be substituted
with the area of the connected member side image that cannot be
photographed from the body trunk side camera for being located
behind the connected member, a composite image that is equivalent
to an image taken by the body trunk side camera without having the
connected member hinder the sight of the body trunk side camera may
be acquired. In most cases, a composite image that equals to the
body trunk side image which has been taken by the body trunk side
camera cannot be achieved merely by replacing the area of the body
trunk side image with the corresponding area of the connected
member side image. This is due to the fact that difference exists
in the position of the body trunk side camera and the connected
member side camera. However, the connected member side camera is
within the direction that cannot be photographed by the body trunk
side camera, and the viewpoint of the body trunk side camera and
the connected member side camera with respect to the area that
cannot be photographed from the body trunk side camera does not
generally differ. Rather, the body trunk side camera and the
connected member side camera may take pictures from substantially
the same direction. Thus, it is possible in some cases to achieve
composite image which equals to an image that was photographed from
the body trunk side camera without the connected member in sight,
by simply changing the scale of the connected member side image and
replacing a part of the body trunk side image with the
corresponding part of the connected member side image of a reduced
(increased in some cases) scale. Furthermore, in a case where
needed, the connected member side camera image may be converted
into an image that is supposed to have been taken by the body trunk
side camera before being composed. The quality of the composite
image can be improved. By utilizing such composite image, location
of the switches existing within the area that cannot be
photographed by the trunk side camera for being behind the
connected member may thus be recognized. It may enable the
connected member to move to the intended switch within such area.
Consequently, the body trunk may be moved towards the intended
switch.
[0017] Another type of configuration may specify an area from the
connected member side image that corresponds to the area in which
the connected member is photographed in the body trunk side image.
With the execution of image processing on the body trunk side
image, a part relevant to the contour of the connected member may
be extracted from the body trunk side image. Such process may be
enabled with a process of extracting feature points from the
surrounding images existing thereby. For example, in a case where
the connected member is positioned in front of a box and the view
of the front plane of the box is partially hindered by the
connected member, feature extraction may be processed on the body
trunk side image to extract the figure that corresponds to the
connected member, the feature points of the parts of the box that
is not hidden by the connected member can be specified. By
specifying feature points that surrounds the area in which the
connected member is photographed, the area including the region in
which the surrounding environment has not been able to be
photographed because of the connected member may be determined. On
the other hand, the aforesaid feature points that surrounds the
area in which the connected member is photographed is also
correspondingly recorded in the connected member side image. Thus,
by extracting the aforesaid feature points from the connected
member side image, an image of the surrounding environment within
the region in which it has not been able to be photographed by the
body trunk side camera because of the connected member may be
determined within the connected member side image. Further by
substituting the former area of the body trunk side image with the
latter area of the connected member side image, a composite image
that equals to an image taken by the body trunk side camera with
the connected arm removed from sight may be obtained.
[0018] As explained above, depending on the environment in which
the robot is operated, there may be a case in which the replacement
area can be specified by calculation, while a case in which the
replacement area can be specified by image processing may also
exist. Note that the two processes described above are mere
examples; the specifying process may not be limited only to the
above described examples.
[0019] In another alternative embodiment of the robot of the
present invention, the movable robot may further comprise a
plurality of body trunk side cameras. In this case, the composite
image creation unit may create composite image for each body trunk
side image taken by the respective body trunk side camera. Under
such a case, a plurality of composite images that are photographed
from different views may be obtained. This may be useful in
specifying the three-dimensional position of the feature point(s)
that exist within the photographed scope in a case, for instance,
where information regarding lateral or depthwise direction is
unknown.
[0020] In another alternative embodiment of the robot of the
present invention, the movable robot may calculate relative
position(s) of one or more objects that exist in the surrounding of
the movable robot by using the composite image. As in the
aforementioned case in which the target of operation is arranged on
a wall surface and the lateral or depthwise information thereof is
given to the robot, the residing position of the operation target
may be determined based on one composite image. Even in a case
where the lateral or depthwise information is unknown to the robot,
the three-dimensional residing position of the surrounding objects
may be determined based on the plurality of composite images that
shows the surrounding objects from different angles.
[0021] In another alternative embodiment of the robot of the
present invention, the movable robot may further comprise a camera
direction adjusting unit that is configured to adjust direction of
the connected member side camera in accordance with the position of
the connected member with respect to the body trunk side camera.
The angle of the area in which the body trunk side camera cannot
photograph with respect to the connected member may differ in
accordance with the position of the connected member. With the
configuration which adjusts the direction of which the connected
member side camera faces in accordance with the position of the
connected member, it may be guaranteed that the area in which the
body trunk side camera cannot photograph is photographed by the
connected member side camera. Occurrence of blind spot may be
effectively prevented.
[0022] In a case where the lateral information is known, valid
information may be obtained by simply creating one composite image.
In a case to the contrary where the lateral information is unknown,
obtainment of image information regarding a plurality of pictures
which are taken from different views and calculation of the
surrounding objects from the obtained image information may be
usefully employed.
[0023] A movable robot that has been created to comply with that
kind of demand may comprise a body trunk, a connected member that
is connected to the body trunk by a joint in which a driving
mechanism is provided, a plurality of body trunk side cameras that
is arranged on the body trunk, a connected member side camera that
is arranged on the connected member, a first calculating unit that
calculates relative positional relationship of one or more objects
existing within a photographed scope based on at least two body
trunk side images taken by one or more of the plurality of body
trunk side cameras and a second calculating unit that calculates
relative positional relationship of one or more objects existing
within a photographed scope based on a body trunk side image taken
by at least one of the plurality of body trunk side cameras and a
connected member side image taken by the connected member side
camera. With the aforementioned configuration, the positional
information is computed by processing image information that may be
obtained from the images. A composite image does not necessarily
have to be created.
[0024] In the aforementioned configuration, a plurality of body
trunk side image information may be obtained by using the plurality
of body trunk side cameras that are each arranged at different
positions. Thereby, the relative positional relationship of the
objects that are photographed in the plurality of body trunk side
images may be calculated with respect to the movable robot.
However, as of the surrounding objects that are not photographed
for being behind the connected member in at least one of the body
trunk side images used for the calculation, the position thereof
cannot be determined.
[0025] With the configuration of the aforementioned movable robot,
it uses a combination of a body trunk side image that is taken by
one of the plurality of body trunk side cameras and a connected
member side image that is taken by the connected member side
camera. This combination of images also provides a plurality of
image information that has been photographed from different views;
the relative positional relationship of the surrounding objects
with respect to the movable robot may be determined.
[0026] Even if the objects cannot be seen from all of the body
trunk side cameras with the connected member obstructing the
sights, so long as the objects are taken by one of the body trunk
side camera, the image thereof may be used in combination with the
information of the connected member side image in the specifying of
the relative positional relationship of the said objects and the
movable robot. According to the aforementioned configuration, the
region remaining unknown to the robot may be lessened.
[0027] In another alternative embodiment of the robot of the
present invention, the second calculating unit may select a body
trunk side image in which portion of the connected member being
photographed is the smallest to be used in the calculation of the
relative positional relationship of the one or more objects with
respect to the movable robot. With the configuration of this robot,
in addition to the cases as illustrated in the unlimiting examples
above, the calculation may be processed by employing the
combinations of information regarding the connected member side
image and respective information regarding each of the body trunk
side images.
[0028] In accordance with the movable robot of the present
invention, the camera arranged on the connected member endows a
supplementary function of visual recognition to the connected
member. The aforementioned configuration contributes in decreasing
or exterminating the unknown regions which cannot be photographed
by the body trunk side camera. With the aforementioned
configuration, the movable robot may promptly and safely be
operated.
BRIEF DESCRIPTION OF THE DRAWINGS
[0029] FIG. 1 shows a perspective view of a robot of the first
embodiment and its surrounding environment.
[0030] FIG. 2 shows a schematic mechanism of the robot of the first
embodiment.
[0031] FIG. 3 shows a view of the robot of the first
embodiment.
[0032] FIG. 4 shows an example of a body trunk side image.
[0033] FIG. 5 shows an example of a connected member side
image.
[0034] FIG. 6 is a flowchart showing a process to be executed by an
image-composition processing device.
[0035] FIG. 7 shows area R in which a connected member is
photographed within the body trunk side image of FIG. 4.
[0036] FIG. 8 shows corresponding area S within the connected
member side image of FIG. 5.
[0037] FIG. 9 shows an example of output of step S613.
[0038] FIG. 10A shows a body trunk side image taken by a body trunk
side camera arranged on the right side of the robot.
[0039] FIG. 10B shows a body trunk side image taken by a body trunk
side camera arranged at the center of the robot.
[0040] FIG. 10C shows a body trunk side image taken by a body trunk
side camera arranged on the left side of the robot.
[0041] FIG. 11 shows another example of a plurality of
corresponding areas S within a connected member side image
[0042] FIG. 12 shows a perspective view of a robot of the second
embodiment and its surrounding environment.
[0043] FIG. 13 shows a schematic mechanism of the robot of the
second embodiment.
[0044] FIG. 14 shows a top plan view of the robot of the second
embodiment
[0045] FIG. 15 is a flowchart showing a process of operation of the
robot of the second embodiment.
[0046] FIG. 16 is a flowchart showing an image processing for
obtaining stereo image.
[0047] FIG. 17 shows a schematic mechanism of the robot of the
third embodiment.
EMBODIMENTS OF THE INVENTION
[0048] Some of the preferred features of embodiments in which the
present invention may be carried out will be listed below.
(Mode 1) The robot may comprise a moving mechanism arranged on the
body trunk. The embodiments disclosed in the present specification
employs wheels, but the moving mechanism may be legged links. (Mode
2) The robot may calculate the course of moving passage of the body
trunk based on an environment map. (Mode 3) The robot may calculate
the moving route of the connected member based on an environment
map. (Mode 4) The robot may calculate the moving route of the body
trunk and the connected member based on an environment map. (Mode
5) The robot may control the attitude of the robot and the
photographing direction of respective cameras by employing joint
angle controller, wheel revolution controller, positional
information detection device and the like. A body trunk side camera
direction calculating unit may calculate the photographing
direction of the body trunk side camera, and a connected member
side camera direction calculating unit may calculate the
photographing direction of the connected member side camera (Mode
6) The robot may create the moving route based on the information
earned from the composite image. The joint angle controller, wheel
revolution controller, positional information detection device
controls actuator(s) to move the body trunk and/or the connected
member in accordance with the created moving route. (Mode 7) The
robot may comprise a route correction unit that makes correction in
accordance with the changes that occur in the surrounding
environment to the moving route (course) that it previously had
made.
First Embodiment
[0049] An unlimiting first embodiment of the present invention will
be described below with reference to the drawings. FIG. 1 shows a
perspective view of a robot 100 of this embodiment and its
surrounding environment. The robot 100 is provided with a body
trunk 103, a connected member (hereinafter referred to as "arm")
104 connected to the body trunk 103 by a shoulder joint 203. A
plurality of arms 104 may be provided. The body trunk 103 includes
a torso 200, a neck joint 202 and head 201. A body trunk side
camera 105 is fixedly arranged on the head 201. The body trunk side
camera 105 changes its photographing direction (hereinafter may
also be referred to simply as "direction" or "camera direction")
with respect to the torso 200 by the rotation of the neck joint
202. A pair of coaxial wheels 206 is arranged on the lower side of
the body 200. The robot 100 moves the body trunk 103 by using the
wheels 206.
[0050] The arm 104 includes a shoulder joint 203, elbow joint 204,
and wrist joint 205. An arm side camera 106 is arranged on the hand
part. The shoulder joint 203, elbow joint 204, and wrist joint 205
are each installed with a driving mechanism. Thus, the arm side
camera 106 may change its direction (photographing direction) with
respect to the body 200 by the driving force of the shoulder joint
203, elbow joint 204, and wrist joint 205. Furthermore, the
position of the arm side camera 106 with respect to the torso 200
may also be changed by the works of those joints.
[0051] The robot 100 is further provided with a device that
measures the position of the head 201 within the working
environment of the robot 100 so that the camera position and the
direction of the body trunk side camera 105 are recognized.
Moreover, with a relevant configuration, the camera position and
the direction of the arm side camera 106 with respect to the body
trunk 103 is also recognized.
[0052] The robot 100 of the present embodiment approaches the
switching panel 10 by using the wheels 206, and operates the group
of switches 12a, 12b, etc. by using the arm 104. In this
embodiment, the position of the wall where the switching panel 10
is arranged is given to the robot 100.
[0053] FIG. 2 shows a schematic mechanism 101 of the robot 100. The
mechanism 101 is used in the robot's visual recognition of the
surroundings. The robot 110 of the present embodiment is provided
with a controller 114. The controller 114 includes a composite
image creation device 110, a moving route creation device 113, a
body trunk side camera position/direction calculation device 111,
and an arm side camera position/direction calculation device 109.
The composite image creation device 110 is configured to have a
body trunk side image information input from the body trunk side
camera 105, an arm side image information input from the arm side
camera 106, a position and direction of the body trunk side camera
105 input from the body trunk side camera position/direction
calculation device 111, and a position and direction of the arm
side camera 106 input from the arm side camera position/direction
calculation device 109. Based on the aforementioned inputs, the
composite image creation device 110 creates a composite image from
the body trunk side image info and the arm side image info. The
moving route creation device 113 creates the further moving route
based on the created composite image, and in accordance with the
created moving route, controls the actuators 115 that are provided
to drive the wheels 26 and the joints. Thus, the moving route
creation device 113 controls the actuators 115 such that the wheels
26, the head 201 and arm 104 respectively in accordance with the
created route. The robot 100 is further provided with a wheel axle
revolution detection device 119 that detects the frequency of
revolution of the wheels 26, a joint angle detection device 117
that detects the joint angle of each of the joints, and a
positional information detection device 121 that is arranged in the
torso 200 and detects the position of the torso 200. The
information that are retrieved by the aforesaid devices are
inputted into the body trunk side camera position/direction
calculation device 111 and the arm side camera position/direction
calculation device 109.
[0054] Since the robot comprises the moving route creation device
113 that creates the course to move by recognizing the surrounding
environment from image information driven from cameras whose
positions and photographing directions are known, the camera
position and the camera directions may be calculated again after
moving along the determined route, which may be followed by the
process of taking pictures may be carried out again. The aforesaid
series of processes may be repeated, which enables the robot to
move autonomously.
[0055] FIG. 3 shows a view of the robot 100. In a case where the
arm 104 of the robot 100 comes into the sight of the body trunk
side camera 105, environmental information regarding the
environment behind the arm 104 cannot be obtained. Specifically, in
the body trunk side image that is taken by the body trunk side
camera 105, while the points A and B of the switching panel 10 of
FIG. 3 will be photographed, the points C and D will not be
photographed. Thus, information of the surrounding object in
regards to the points C and D remains unknown, thus has
conventionally been causing problems as to the robot not being able
to sufficiently recognize the environmental circumstance nor to
understand the relative positional relationship of the surrounding
objects with respect to the robot 100. In contrast, the visible
range is enlarged by the application of the arm side camera 106
with the technique of the present embodiment; the present technique
enables the robot to recognize the unknown area including the
points C and D.
[0056] FIGS. 4 and 5 respectively shows an example of the body
trunk side image and the arm side image. Though the panel is
partially hidden by the arm 104 in the body trunk side image of
FIG. 4, the corresponding surface of the panel is recorded in the
arm side image of FIG. 5. Such relationship of the body trunk side
image and the arm side image is established with the position of
the head 201 and the arm 104 being controlled with the detection of
the joint angle detection device 117, and the camera direction of
the arm side camera 106 being adjusted in accordance with the
relative position of the body trunk side camera 105 and the arm
104. The camera direction of the arm side camera 106 is adjusted
such that the arm side camera 106 photographs in the direction in
which the area that is hidden from the body trunk side camera by
the arm 104 may be photographed.
[0057] Information regarding magnification of the camera, image
size and the like of the respective cameras are obtained by either
the body trunk side camera position/direction calculation device
111 and the arm side camera position/direction calculation device
109. Furthermore, since image processing based on the aforesaid
information will be carried out at the time of composing the images
to create a new image, the photographed scope of each image does
not have to be of a complete match. For example, The distance
between the switching panel 10 and the body trunk side camera 105
may differ from that of the switching panel 10 and the arm side
camera 106, and the arm 104 may be tilted with respect to the angle
of the head 201. Due to the aforementioned reasons, the arm side
image of FIG. 5 depicts the switching panel 10 in an enlarged
magnitude and also angled with respect to the body trunk side image
of FIG. 4.
[0058] FIG. 6 shows a flowchart of the image composite process. In
the explanation below, it is assumed that the robot 100 and the
switching panel 10 are in a positional relationship as shown in
FIG. 1 at the time of starting the below process. Simultaneously,
the arm 104 of the robot 100 is at a position as shown in FIG. 4,
and such image as in FIG. 4 is photographed by the body trunk side
camera 105. In the below-assumed situation, the robot 100 is trying
to move the arm 104 in order to operate switch 12c (see FIG. 1),
however, the target switch 12c is within the unrecognizable area
within the body trunk side image for being behind the arm 104. The
process of FIG. 6 may be commenced under such a condition.
[0059] In step S601, information regarding the body trunk side
image is input into the composite image creation device 110. In
step S603, feature points X.sub.n of the unknown environment that
is unknown within the body trunk side image are extracted. FIG. 7
shows an example of the extracted feature points. Since the contour
of the arm 104 is known to the robot 100, the area in which the arm
104 is photographed (which is a photographed scope of the arm) can
be specified from the body trunk side image. This equals to the
employment of a first specifying unit. The points X.sub.1-X.sub.6
of the figure are feature points that were extracted to surround
the photographed scope of the arm. In the present embodiment, six
individual feature points that can be linearly connected are
extracted. However, the feature points X.sub.n may be extracted in
any form so long as they are aligned along or in accordance with
the contour of the arm. In step S605, the area R in which the arm
104 is photographed (that is, the area whose surrounding object
information is unknown) is specified from the feature points that
have been extracted in step S603. The area of FIG. 7 that is
bordered by the feature points X.sub.1-X.sub.6 is the area R. The
area R of the present embodiment roughly resembles the outline of
the arm 104. The area R that is bordered by the feature points may
closely resemble the outline of the arm in a more genuine manner by
increasing the number of feature points X.sub.n to be
extracted.
[0060] In step S607, information regarding the arm side image is
input into the composite image creation device 110. In step S609,
feature points Z.sub.n that corresponds to the feature points that
defines the area R of the body trunk side image are extracted. FIG.
8 shows an example of the extracted feature points. The feature
points Z.sub.1-Z.sub.6 of FIG. 8 are the feature points that
correspond to the feature points X.sub.1-X.sub.6 extracted of the
body trunk side image. In step S611, the corresponding area S (an
area that corresponds to the unknown area of the body trunk side
image) is specified from the feature points that have been
extracted in step S609.
[0061] In step S613, adjustment of size, rotation of angle and the
like is processed, based on the difference information of the input
images from the body trunk side camera and the arm side camera, on
the corresponding area S that is obtained from the arm side image.
The adjustment process is carried out so that the corresponding
area S can appropriately be applied to the area R that depicts the
arm within the body trunk side image.
[0062] In step S615, the area R of the body trunk side image is
replaced with the corresponding area S, of which had gone through
the aforementioned process to prepare for the image composition. As
the result of the replacement process, a composite image which
unknown area has been excluded from view is outputted in step S617.
FIG. 9 shows an example of such composite image. The image within
the area R shown with broken like has been substituted with the
image of area S. Thereby, a composite image in which the arm is not
included can be obtained.
[0063] The robot of the present embodiment may be provided with a
plurality of body trunk side cameras. In such a case, the area
R.sub.x that depicts the arm within each of the body trunk side
image may be extracted from respective body trunk side image, and
corresponding areas S.sub.x may be extracted from the arm side
image to correspond with each of the extracted area R.sub.x, and
replace each of the areas R.sub.x with the respective corresponding
area S.sub.x.
[0064] FIG. 10 shows examples of areas R of the body trunk side
images. In a case where the robot is arranged with three body trunk
side cameras; one on left and right side each and another at the
center of the body trunk 103, a body trunk side image TC.sub.i as
shown in FIG. 10A may be obtained by the body trunk side camera on
the right side, a body trunk side image TC.sub.2 as shown in FIG.
10B may be obtained by the body trunk side camera at the center
position, and a body trunk side image TC.sub.3 as shown in FIG. 10C
may be obtained by the body trunk side camera on the left side. The
body trunk side images TC.sub.1, TC.sub.2, TC.sub.3 go through the
process of steps S601 to S605, and area R.sub.1, R.sub.2, R.sub.3
have been specified in each images respectively. Due to the
difference in the position of which each of the body trunk side
cameras are arranged, the position of which the arm is photographed
differs in the body trunk side images TC.sub.1, TC.sub.2,
TC.sub.3.
[0065] FIG. 11 shows an example of arm side image taken by the arm
side camera that operates in accordance with the body trunk side
cameras. In order to supplement information of the unknown areas
TC.sub.1, TC.sub.2 and TC.sub.3, process of steps S607 to S611 are
carried out on the arm side image AC to specify the corresponding
areas S.sub.1, S.sub.2 and S.sub.3.
[0066] With the process of steps S613 to S617, the areas R.sub.1,
R.sub.2 and R.sub.3 of the respective body trunk side cameras
TC.sub.1, TC.sub.2 and TC.sub.3 are replaced with respective
corresponding area S.sub.1, S.sub.2 and S.sub.3 of the arm side
image. As a result, a plurality of body trunk side images whose
unknown areas has been supplemented can be obtained.
[0067] Furthermore, the robot 100 of the present embodiment may be
provided with a plurality of arm side cameras. In such a case, it
is preferable that a determination device that determines and
selects for each of the area R of the body trunk side images an arm
side image in which a corresponding area S that most widespreadly
covers the area R is specified.
[0068] The robot 100 of the present embodiment may further
calculate the relative position of the surrounding objects with
respect to the robot 100 by using the composite image(s) obtained
by the aforementioned procedures. The relative position of the
surrounding objects with respect to the robot 100 may be recognized
by converting the relative positional coordinates of the object in
the world coordinate system into coordinates within a local
coordinate system whose center is set at the location of the
robot.
Second Embodiment
[0069] As for the configurations in common with the first
embodiments, the same reference numericals are given and the
explanation thereof may be omitted.
[0070] FIG. 12 shows a perspective view of a stereo camera robot of
this embodiment and its surrounding environment. The robot 100 of
the present embodiment is provided with a body trunk 103, and a
connected member (hereinafter referred to as "arm") 104 connected
to the body trunk 103 by a shoulder joint 203. A plurality of arms
104 may be provided. The body trunk 103 includes a torso 200, a
neck joint 202 and head 201. A pair of body trunk side cameras
105(a) and 105(b) are fixedly arranged on the head 201. The body
trunk side cameras 105(a) and 105(b) change their photographing
directions (hereinafter may be referred to simply as "directions"
or "camera directions") with respect to the torso 200 by the
rotation of the neck joint 202. A pair of coaxial wheels 206 is
arranged on the lower side of the torso 200. The robot 100 moves
the body trunk 103 by using the wheels 206.
[0071] The arm 104 includes a shoulder joint 203, elbow joint 204,
and wrist joint 205. An arm side camera 106 is arranged on the hand
part. The shoulder joint 203, elbow joint 204, and wrist joint 205
are each installed with a driving mechanism. Thus, the arm side
camera 106 may change its camera direction with respect to the
torso 200 by the driving force of the shoulder joint 203, elbow
joint 204, and wrist joint 205. Furthermore, the position of the
arm side camera 106 with respect to the torso 200 may also be
changed by the works of those joints.
[0072] The robot is further provided with a device that measures
the position of the head 201 within the working environment of the
robot 100 so that the camera positions and the directions of the
body trunk side cameras 105(a) and 105(b) are recognized. Moreover,
with a relevant configuration, the camera position and the
direction of the arm side camera 106 with respect to the body trunk
103 is also recognized.
[0073] The robot 100 of the present embodiment visually recognizes
the surrounding objects 30, 40, 50 and 60, and plans a route in
which it will not collide with those obstacles and operates to
place the holding object 20 in an empty space. The body trunk side
cameras 105(a) and 105(b) generally functions as a stereo camera.
The points A, B, C and D in FIG. 12 indicates examples of focusing
point of the robot's vision.
[0074] FIG. 13 shows a schematic mechanism 101 of the robot 100.
The mechanism 101 is used in the robot's visual recognition of the
surroundings. The robot 110 of the present embodiment is provided
with a controller 114. The controller 114 includes an image
information processing device 116, an environment map storage
device 112, a moving route creation device 113, a body trunk side
camera position/direction calculation device 111, and an arm side
camera position/direction calculation device 109. The image
information processing device 116 is configured to have body trunk
side image information input from the body trunk side cameras
105(a) and 105(b), an arm side image information input from the arm
side camera 106, positions and directions of the body trunk side
cameras 105(a) and 105(b) input from the body trunk side camera
position/direction calculation device 111, and a position and
direction of the arm side camera 106 input from the arm side camera
position/direction calculation device 109. Based on the
aforementioned inputs, the image information processing device 116
creates an environment map and stores the created environment map
in the environment map storage device 112. The moving route
creation device 113 creates the further moving route based on the
created environment map, and in accordance with the created moving
route, controls the actuators 115 that are provided to drive the
wheels 26 and the joints. Thus, the moving route creation device
113 controls the actuators 115 such that the wheels 26, the head
201 and arm 104 to move respectively in accordance with the created
route. The robot 100 is further provided with a wheel axle
revolution detection device 119 that detects the frequency of
revolution of the wheels 26, a joint angle detection device 123
that detects the joint angle of each of the joints, and a
positional information detection device 121 that is arranged in the
torso 200 and detects the position of the torso 200. The
information that are retrieved by the aforesaid devices are
inputted into the body trunk side camera position/direction
calculation device 111 and the arm side camera position/direction
calculation device 109.
[0075] Since the robot comprises the moving route creation device
113 that creates the course to move by recognizing the surrounding
environment from image information driven from cameras whose
positions and photographing directions are known, the camera
position and the camera directions may be calculated again after
moving along the determined route, which may be followed by the
process of taking pictures may be carried out again. The aforesaid
series of processes may be repeated, which enables the robot to
move autonomously.
[0076] FIG. 14 shows a top plan view of the sight of the robot 100.
As are shown, if the arm 104 of the robot 100 and its holding
object 20 is in the sight of the body trunk side cameras 105(a) and
105(b), environmental information regarding the are behind the arm
104 and its holding object 20 cannot be obtained. Specifically, as
shown in FIG. 14, the points A and B are visible to the body trunk
side cameras 105(a) and 105(b) as shown with solid lines, thus
stereo image may be created by using body trunk images taken by the
body trunk side cameras 105(a) and 105(b). Moreover, information
regarding those areas can be serially obtained in accordance with
the progress of time.
[0077] On the other hand, as of the points C and D, a full vision
thereof from both of the body trunk side cameras 105(a) and 105(b)
cannot be established due to the operation of the arm or the hand
part of the arm, thereby resulting in the stereovision of the
points C and D not being possible with the body trunk side cameras
105(a) and 105(b). Information regarding those areas cannot be
obtained. The broken lines indicate that an image including the
feature point at focus cannot be photographed by both cameras due
to the obstruction of the arm or its holding object.
[0078] In the case of stereovision, especially in a case where
information regarding the lateral width (depth) or the height of
surrounding objects are necessary, disparity information based on a
plurality of images must be obtained. The plurality of images may
be obtained either by taking pictures using a plurality of cameras
whose positional information are known to the robot, or using a
single camera and take pictures at different positions by moving
the camera. Thus, due to the condition that the robot cannot make
focus on either of the points C and D with both eyes (the body
trunk side cameras), the positional relationship that lies in the
areas of and around the points C and D cannot be recognized.
[0079] As in the case with points C and D where at least one of the
body trunk side cameras 105(a) and 105(b) cannot function as stereo
camera, a stereo image may be obtained by one of the body trunk
side cameras 105(a) and 105(b) and the arm side camera 106.
[0080] In a case where nothing obstructs the view of the body trunk
side cameras 105(a) and 105(b), as it is in the case of points A
and B, the stereo image can be obtained by using the body trunk
side cameras 105(a) and 105(b) and calculate the relative
positional relationship of the surrounding object and the robot
100. This equals to the employment of a first calculating unit. On
the other hand, in a case as it is in the case of points C, the
stereo image can be obtained by using the body trunk side camera
105(b) and the arm side camera 106, while in a case as it is in the
case of points D, the stereo image can be obtained by using one of
the body trunk side camera 105(a) and the arm side camera 106.
Information that is necessary to recognize the surrounding
environment on the whole can be obtained by the aforementioned
camera combinations, and the robot may calculate the relative
positional relationship of the surrounding object and the robot
100. This equals to the employment of a second calculating
unit.
[0081] In a case of using the arm side camera 106 as one of the
stereo camera, the selection of the other camera from the body
trunk side cameras 105(a) and 105(b) may be determined such that a
body trunk side camera whose area in which the arm 104 to be
photographed is the smallest is selected. The image information
processing device 116 may execute the selection.
[0082] FIG. 15 shows a flowchart of a process of operation of the
robot of the present embodiment. The process below will be
described in accordance with a case in which the robot 100 holds
the object 20, and creates a route to place the holding object 20
in an empty space without colliding with the surrounding objects
30, 40, 50 and 60. The arm 104 of the robot 100 is located at a
position as shown in FIG. 14, and the situation as shown in that
figure is photographed by the body trunk side cameras 105(a) and
105(b) from where they are located. The image information
processing as shown in FIG. 15 is commenced, for the arm 104 and
the holding object 20 hide part of the objects 50 and 60 and thus
the hidden part remains unknown to the robot.
[0083] In step S500, a stereo image is obtained from a plurality of
images taken by the body trunk side cameras 105(a) and 105(b) and
the arm side camera 106 that is input to the image information
processing device 116. The image information processing device 116
may determine whether or not to use the arm side image in
accordance with the size of the unknown areas within the body trunk
side images. In such a case, the image information processing
device 116 may comprise a mechanism that obtains stereo image from
a combination of images in which the area that is obstructed by the
arm 104 and become unknown is less photographed.
[0084] In step S502, based on the stereo image obtained in step
S500, the relative positional relationship of the objects
surrounding the robot 100 is calculated. The image data is
converted from the world coordinate system to the local coordinate
system in which the robot 100 is placed at the center. The
positional relationship of the surrounding objects are calculated
in accordance with the local coordinate system, and outputted as
relative positional information of the robot 100 and its
surrounding objects. In the process of acquiring the relative
positional information, for example, disparity images or distance
images may be utilized.
[0085] Furthermore, it is preferable that a first calculating
device and a second calculating device are respectively provided
for calculation in a case of obtaining stereo image from a
combination of body trunk side images and for calculation in a case
of obtaining stereo image from a combination of body trunk side
image and arm side image. As for the combination of body trunk side
images, the body trunk side cameras are parallel stereo. However,
as for the combination of body trunk side image and arm side image,
the relative positions of the cameras differ. Thus, the second
calculating device may be configured to apply coordinate system
converting process, such as rectification of stereo image, hand-eye
calibration, or the like.
[0086] In step S504, the relative positional information calculated
by the image information processing device 116 is constructed as
data that express the relative positioning of the surrounding
objects and stored in the environment map storage device 112. An
environment map refers to data of which the robot 100 uses to
recognize the relative positioning of the surrounding objects. For
example, in a case of computing a disparity image with the image
information processing device 116, the environment map may be
expressed by a spatial disparity space (SDS). In the SDS, the
surrounding objects 30, 40, 50 and 60 including the points A, B, C
and D appear as curved surfaces having lateral or depthwise spatial
expansion.
[0087] In step S506, a moving route of the robot 100 is created
based on the environment map information. Since the robot 100 of
the present embodiment aims to visually recognize the surrounding
objects 30, 40, 50 and 60, create a moving route that will not
collide with those obstacles, and place the holding object 20 in an
empty space, it is preferable that the robot 100 extracts plane(s)
in which the surrounding object does not exist based on such
environmental information, and create route for the arm to place
the holding object 20 within the extracted plane. In step S508, the
actuators 115 operate the actual motion.
[0088] FIG. 16 is a flowchart showing an image processing for
obtaining stereo image. As aforementioned, the stereo image in step
S500 of FIG. 15 may be obtained basically by using a plurality of
body trunk side images, or using one of the body trunk side images
and an arm side image. In step S160, a plurality of body trunk side
images taken by the body trunk side cameras 105(a) and 105(b) and
an arm side image taken by the arm side camera 106 are input into
the image information processing device 116. In step S162, it is
determined whether or not the arm 104 is photographed within the
body trunk side images. In a case where the arm 104 is not
photographed in the body trunk side images ("NO" to step S162), the
process proceeds to step S164. In step S164, a stereo image is
obtained from a pair of body trunk side images, and the first
calculation is processed. On the other hand, in a case where the
arm 104 is photographed in the body trunk side images ("YES" to
step S162), the process proceeds to step S166. In step S166, a
stereo image is obtained from a pair of one of the body trunk side
images and the arm side image, and the second calculation is
processed. The body trunk side image that is selected in step S166
is the body trunk side image in which the area the arm 104 is
photographed is the smallest.
[0089] In the present embodiment, an example of utilizing a total
of three cameras arranged on the body trunk and the connected
member to obtain a three-dimensional information of the surrounding
objects has been indicated. However, the embodiment of the present
invention is not limited to such configuration. It may utilize
larger number of cameras, or may simply use a connected member side
camera and obtain photographs in the series of moving the connected
member side camera and changing the photographing angle or
position.
[0090] Furthermore, the present embodiment has given an example of
recognizing surrounding objects that are reposed to be still.
However, with the aforementioned configuration, the robot may not
be able to cope with a situation in which a new surrounding object
has come to appear in the moving course of the robot after the
robot is set in motion in accordance with the previously created
moving route, or a situation in which the positions of the
surrounding object and the robot have changed due to the motions
carried out by the wheels or the connected member, and the moving
route needs to be corrected.
Third Embodiment
[0091] FIG. 17 shows a schematic mechanism 101 of the robot of the
third embodiment that is configured to deal with surrounding
objects in motion. The mechanism 101 of the robot is, in addition
to the configuration of the mechanism of the robot of the second
embodiment, further provided with a moving route correction device
125. The moving route correction device 125 makes correction to the
created moving route of the connected member. Thus, the robot is
able to dynamically correct its moving course which has already
been put into process.
[0092] Furthermore, since the surrounding environmental data may
dynamically be updated, the change in the surrounding environment
and the motion of the arm may be serially recognized, and the
moving course may be corrected in accordance with the change.
[0093] The specific embodiment of the present invention is
described above, but these merely illustrate some embodiments of
the invention and do not restrict the claims thereof. The art set
forth in the claims includes various transformations and
modifications to the specific embodiments as set forth above.
[0094] For example, aside from the arm side camera arranged on the
arm, one or more connected member side camera(s) may be arranged at
a plurality of connected members; the legs or the like. Or, a
plurality of cameras may be rotatably arranged at different
positions of the connected member. In either of the aforesaid
examples, photographable scope may further be expanded.
[0095] The technical elements disclosed in the present
specification or figures may be utilized separately or in all types
of conjunctions and are not limited to the conjunctions set forth
in the claims at the time of filing of the application.
Furthermore, the art disclosed in the present specification or
figures may be utilized to simultaneously realize a plurality of
aims or to realize one of these aims.
* * * * *